810-110 AITECH — Cisco AI Technical Practitioner Cheat Sheet

Last revised: September 16, 2026

Compact Cisco 810-110 AITECH Cheat sheet covering AI/ML concepts, generative AI, data, infrastructure, security, and operations decisions.

Use the tables for a quick pre-exam check. Expand a topic’s notes for explanations, examples, and additional distinctions.

Scope and study context

For Cisco-style AI scenarios, do not stop at “which model is best.” Also identify:

Business outcome and risk tolerance.
Data type, quality, sensitivity, and ownership.
AI approach: rules, ML, deep learning, generative AI, RAG, or agentic workflow.
Placement: edge, campus, branch, data center, cloud, or hybrid.
Network, security, observability, and operations impact.
Governance, monitoring, rollback, and continuous improvement needs.

Use this independent Cheat Sheet to prepare for the Cisco AI Technical Practitioner (810-110 AITECH) exam from Cisco, official exam code 810-110 AITECH. It is designed for the final review stage: confirming the concepts, decision rules, and common traps you should understand before moving into topic drills, mock exams, and detailed explanations.

This page is not affiliated with Cisco and does not replace Cisco’s official exam information. It focuses on practical exam-prep review: AI fundamentals, data handling, model behavior, generative AI, AI infrastructure, networking considerations, security, governance, and operations.

Use this Cheat Sheet first, then move into IT Mastery practice:

Start with topic drills for AI fundamentals, data, model evaluation, GenAI, infrastructure, security, and operations.
For every missed item, classify the miss:
- Concept gap.
- Metric or terminology confusion.
- Scenario decision error.
- Security/governance oversight.
- Misread wording.
Read the detailed explanations, not just the correct option.
Rework missed questions after a delay.
Finish with mixed mock exams to practice switching topics under time pressure.

A strong next step is to choose your weakest area from the checklist above and complete a focused set of original practice questions before attempting a full-length mixed review.

AI, ML, and Generative AI Terms

Term	Exam-ready meaning	Common trap
Artificial intelligence	Systems that perform tasks associated with human intelligence, such as perception, reasoning, prediction, or generation	AI is broader than machine learning
Machine learning	AI technique where models learn patterns from data	ML is not always generative
Deep learning	ML using multi-layer neural networks	Requires more data/compute than many classical models
Generative AI	AI that creates text, code, images, audio, or other outputs	Output may be plausible but false
Foundation model	Large model trained on broad data and adaptable to many tasks	Not automatically safe, private, or accurate
Large language model	Foundation model optimized for language tasks	Does not “know” truth; predicts likely tokens
Token	Unit of text processed by a model	Token count affects context, latency, and cost
Context window	Maximum input/output tokens the model can consider in one interaction	Larger context does not guarantee better reasoning
Embedding	Numeric vector representing semantic meaning	Embeddings support similarity search, not exact truth
Vector database/index	Stores and searches embeddings by similarity	Retrieval quality depends on chunks, embeddings, metadata, and ranking
Inference	Using a trained model to produce predictions or outputs	Different from training or fine-tuning
Training	Learning model parameters from data	Usually more expensive and data-intensive than inference
Fine-tuning	Additional training on task/domain examples	Not the default fix for missing facts
Prompt engineering	Designing instructions and context for model output	Prompting cannot fully replace controls and evaluation
RAG	Retrieval-augmented generation; retrieves external content and injects it into prompt context	RAG can still hallucinate if retrieval or grounding is poor
Agentic AI	AI system that plans steps and uses tools/APIs to complete tasks	Tool authorization and auditability are critical
Hallucination	Confident but incorrect or unsupported output	Reduced by grounding, evaluation, and guardrails; not eliminated
Model drift	Degradation as real-world data changes	Requires monitoring and retraining/update strategy
Bias	Systematic unfairness or skew in data/model outputs	Can exist even with high overall accuracy
Explainability	Ability to understand why a model made a decision	More important in regulated, high-risk, or user-impacting decisions

AI Workload Lifecycle

    flowchart LR
	    A[Define business problem] --> B[Identify data sources]
	    B --> C[Prepare and label data]
	    C --> D[Choose AI approach]
	    D --> E[Train, configure, or prompt]
	    E --> F[Evaluate]
	    F --> G[Deploy]
	    G --> H[Monitor]
	    H --> I[Improve or retire]
	    I --> B

Notes and examples

Phase	Key questions	Exam focus
Problem framing	Is the goal prediction, classification, generation, search, summarization, automation, or anomaly detection?	Match AI method to outcome
Data sourcing	What data is available, trusted, labeled, current, and permitted for use?	Data quality and governance
Preparation	Clean, normalize, transform, label, split, deduplicate, and protect data	Avoid leakage and bias
Model or approach selection	Rules, classical ML, deep learning, LLM, RAG, fine-tuning, or agent?	Choose the simplest approach that meets requirements
Evaluation	What metric reflects the business risk?	Accuracy is not always enough
Deployment	Where should inference run? What latency, cost, privacy, and connectivity constraints exist?	Edge/data center/cloud tradeoffs
Operations	How will drift, failure, abuse, and infrastructure issues be detected?	MLOps/LLMOps and observability

Choosing the Right AI Approach

Scenario	Prefer	Why	Watch for
Known rules, stable logic, low ambiguity	Rules or automation	Deterministic, auditable, simpler	Do not add ML where rules are enough
Predict a numeric value	Supervised regression	Learns relationship between features and continuous target	Outliers and changing data distributions
Assign item to known category	Supervised classification	Learns from labeled examples	Class imbalance and false positive/negative cost
Discover groups without labels	Clustering	Finds natural groupings	Clusters may not map to business categories
Detect unusual behavior	Anomaly detection	Useful for operations, security, fraud, performance	Needs baseline and tuning
Optimize actions through feedback	Reinforcement learning	Learns policies from rewards	Can be complex and risky in production
Understand images/video	Computer vision / CNNs / vision transformers	Extracts visual patterns	Data labeling and edge compute requirements
Summarize or generate text	LLM / generative AI	Produces natural-language output	Hallucination, privacy, prompt injection
Answer using enterprise documents	RAG	Grounds answers in retrieved content	Poor chunking/retrieval causes wrong answers
Adapt tone, format, or domain behavior	Prompting or fine-tuning	Prompt first; fine-tune when examples are needed	Fine-tuning does not reliably add fresh knowledge
Execute multi-step tasks with tools	Agentic workflow	Plans, calls APIs, and uses memory/tools	Tool permissions, audit logs, and guardrails

Model and Algorithm Selection

Model/technique	Best fit	Strengths	Limitations/traps
Linear regression	Numeric prediction with linear relationships	Simple, explainable, fast	Poor for complex nonlinear patterns
Logistic regression	Binary or multiclass classification	Interpretable baseline	Name says regression but used for classification
Decision tree	Classification/regression with explainability	Easy to visualize	Can overfit
Random forest	General tabular prediction	Reduces overfitting vs single tree	Less interpretable
Gradient boosting	High-performing tabular prediction	Strong accuracy on structured data	Tuning and overfitting risk
k-nearest neighbors	Similarity-based classification/regression	Simple concept	Slow at scale; sensitive to feature scaling
Support vector machine	Classification with clear margins	Effective in some high-dimensional spaces	Less transparent; scaling concerns
k-means	Clustering into predefined number of groups	Fast and common	Must choose k; assumes roughly spherical clusters
DBSCAN	Density-based clustering	Finds irregular clusters and noise	Sensitive to parameters and density variation
Isolation forest	Anomaly detection	Good for outlier discovery	Needs tuning and validation
Neural network	Complex nonlinear patterns	Flexible	Requires more data, compute, and monitoring
CNN	Images/spatial data	Strong visual feature extraction	Training data and compute intensive
RNN/LSTM	Sequential/time-series data	Captures sequence patterns	Often replaced by transformer approaches for many language tasks
Transformer	Language, multimodal, sequence tasks	Parallelizable, strong context modeling	Compute, latency, and governance concerns
LLM	Text generation, summarization, reasoning assistance	Flexible natural-language interface	Probabilistic and may hallucinate

Notes and examples

Model selection quick rules

If the problem asks for…	Likely approach	Notes
Yes/no or category prediction	Supervised classification	Evaluate with precision, recall, F1, ROC-AUC, or PR-AUC.
Numeric prediction	Supervised regression	Evaluate with MAE, RMSE, R-squared, business error tolerance.
Group similar items	Clustering	No labels required; evaluation can be harder.
Find unusual behavior	Anomaly detection	Useful for fraud, security, sensor faults, network anomalies.
Recommend items	Recommendation/ranking	Consider user behavior, feedback loops, fairness, cold start.
Image recognition	Computer vision / CNNs / vision transformers	Data quality and labeling are critical.
Natural language understanding	NLP models or LLMs	Consider embeddings, context, latency, and hallucination risk.
Generate text or code	LLM / GenAI	Use prompts, RAG, guardrails, and evaluation.
Make sequential decisions with rewards	Reinforcement learning	More complex; environment and reward design matter.

Classical ML vs deep learning vs LLMs

Choice	Strengths	Limitations	Best fit
Classical ML	Often explainable, efficient, strong on tabular data.	May need feature engineering; limited for unstructured language/image tasks.	Structured business data, smaller datasets, low-latency needs.
Deep learning	Strong for images, speech, text, complex patterns.	Data-hungry, compute-intensive, harder to explain.	Unstructured data and large-scale pattern recognition.
Pre-trained foundation model	Fast adaptation, broad capabilities.	Cost, latency, privacy, hallucination, governance concerns.	Summarization, Q&A, content generation, semantic search.
Custom model from scratch	Maximum control.	Expensive, data-intensive, operationally complex.	Specialized needs with sufficient data, budget, and expertise.

Data Reference

Data Types

Data type	Examples	AI handling notes
Structured	Tables, logs with fields, CRM records, metrics	Common for classical ML and analytics
Semi-structured	JSON, XML, event records, emails with metadata	Requires parsing and schema management
Unstructured	Text, images, audio, video, PDFs	Often needs embeddings, OCR, NLP, CV, or multimodal models
Time-series	Telemetry, sensor data, network metrics	Preserve time order; avoid random leakage
Graph data	Network topology, identity relationships, dependencies	Useful for relationship and path analysis
Streaming data	Live telemetry, alerts, events	Requires low-latency processing and monitoring

Notes and examples

Data Preparation Decisions

Task	Purpose	Exam trap
Cleaning	Remove errors, duplicates, invalid values	Do not silently remove meaningful outliers
Normalization/scaling	Put numeric features on comparable scales	Especially important for distance-based methods
Encoding	Convert categories to numeric representation	Ordinal vs one-hot encoding matters
Tokenization	Break text into model-processable units	Token limits affect prompt design
Chunking	Split documents for retrieval	Chunks too small lose context; too large reduce precision
Labeling	Add target values/classes	Label quality often limits model quality
Train/validation/test split	Separate training, tuning, and final evaluation	Never tune on the test set
Cross-validation	Repeated train/validate cycles	Useful with limited data
Feature engineering	Create useful inputs from raw data	Can introduce data leakage
Deduplication	Remove repeated records	Duplicates across train/test inflate scores
Class balancing	Address rare classes	Accuracy can hide poor minority-class performance
Data masking	Protect sensitive data	Mask before using data in prompts or training when required

Data lifecycle review

AI projects usually fail from poor data, weak evaluation, or operational mismatch before they fail from lack of model sophistication.

    flowchart LR
	    A[Define use case] --> B[Collect and govern data]
	    B --> C[Clean, label, and validate]
	    C --> D[Split data]
	    D --> E[Train or select model]
	    E --> F[Evaluate]
	    F --> G[Deploy]
	    G --> H[Monitor]
	    H --> I[Retrain or adjust]
	    I --> C
	    B --> J[Security, privacy, lineage]
	    F --> J
	    G --> J
	    H --> J

Data concepts to review

Concept	Why it matters	Candidate mistake
Training data	Used to fit the model.	Letting test data influence training decisions.
Validation data	Used for model selection and tuning.	Treating validation results as final unbiased performance.
Test data	Held out for final evaluation.	Reusing the test set repeatedly until it becomes part of tuning.
Labels	Known answers for supervised learning.	Assuming labels exist or are reliable.
Features	Input variables used by the model.	Including leakage features that reveal the target indirectly.
Data leakage	Information from the future or target slips into training.	Getting unrealistically high test scores that fail in production.
Imbalanced data	One class is much more common than another.	Using accuracy when precision/recall is more relevant.
Data drift	Input data distribution changes over time.	Assuming a model remains valid forever.
Concept drift	Relationship between inputs and target changes.	Monitoring only data shape, not outcome quality.
Lineage	Traceability of where data came from and how it changed.	Being unable to explain, audit, or reproduce a model result.

Data splitting trap

A random split is not always appropriate.

Scenario	Better split approach
Time-series forecasting	Train on earlier time periods, validate/test on later periods.
User-level behavior	Split by user/account/entity to avoid leakage.
Medical, finance, or regulated data	Preserve privacy, auditability, and representative sampling.
Rare events	Use stratified splitting where appropriate.
Duplicated or near-duplicated records	Deduplicate or group before splitting.

Evaluation Metrics

Use case	Metric	Meaning	When to prefer
Balanced classification	Accuracy = correct / total	Overall correctness	Classes are balanced and errors have similar cost
Imbalanced classification	Precision = TP / (TP + FP)	How many predicted positives are correct	False positives are costly
Imbalanced classification	Recall = TP / (TP + FN)	How many actual positives are found	False negatives are costly
Classification tradeoff	F1 = 2 × precision × recall / (precision + recall)	Balances precision and recall	Need one score for imbalanced classification
Threshold evaluation	ROC-AUC	Separability across thresholds	Compare classifiers independent of one threshold
Regression	MAE	Average absolute error	Easy to interpret
Regression	MSE/RMSE	Penalizes larger errors	Large mistakes are especially costly
Regression	R-squared	Variance explained	Not enough alone for operational fit
Clustering	Silhouette score	Cluster separation/cohesion	Internal clustering assessment
Ranking/retrieval	Top-k accuracy, MRR, NDCG	Quality of ranked results	Search, recommendation, RAG retrieval
Generative AI	Faithfulness/groundedness	Output supported by provided source	RAG and enterprise Q&A
Generative AI	Toxicity/safety score	Harmful or unsafe content risk	Public-facing or user-impacting systems
Operations	Latency, throughput, error rate	Runtime performance	Production readiness
Business	Cost per task, containment rate, analyst time saved	Outcome value	Executive and operational alignment

Notes and examples

Confusion Matrix Terms

Term	Meaning	Example
True positive	Predicted positive and actually positive	Correctly flagged threat
False positive	Predicted positive but actually negative	Benign activity flagged as threat
True negative	Predicted negative and actually negative	Correctly ignored benign activity
False negative	Predicted negative but actually positive	Missed actual threat

High-yield trap: A model can have high accuracy but poor recall if the positive class is rare.

Generative AI and RAG

RAG Pipeline

    flowchart LR
	    A[Source documents] --> B[Clean and split into chunks]
	    B --> C[Create embeddings]
	    C --> D[Store in vector index]
	    E[User question] --> F[Embed question]
	    F --> G[Retrieve relevant chunks]
	    G --> H[Optional rerank/filter]
	    H --> I[Prompt with context]
	    I --> J[Generate answer]
	    J --> K[Evaluate, cite, log]

Notes and examples

RAG component	Purpose	Failure mode
Source selection	Choose trusted knowledge	Inaccurate or outdated sources cause bad answers
Chunking	Create retrievable units	Bad chunk size loses context or lowers precision
Embedding model	Convert text to vectors	Poor semantic match reduces retrieval quality
Vector index	Store and search embeddings	Missing metadata/filtering returns irrelevant content
Retriever	Finds candidate chunks	Low recall misses needed facts
Reranker	Improves ordering of results	Adds latency
Prompt template	Combines task, rules, and retrieved context	Weak instructions allow unsupported answers
Generator	Produces final response	May hallucinate if context is weak
Guardrails	Enforce safety, format, and policy	Overly broad guardrails block useful responses
Evaluation set	Tests answer quality	No eval set means quality changes go unnoticed

Prompting, RAG, Fine-Tuning, or Training?

Need	Best first choice	Why
Change response format or tone	Prompt engineering	Fastest and lowest complexity
Answer from current enterprise knowledge	RAG	Keeps knowledge external and updateable
Reduce hallucination with citations	RAG plus grounding rules	Model can reference retrieved sources
Teach a model specialized style or repeated task pattern	Fine-tuning	Learns behavior from examples
Add private facts that change often	RAG, not fine-tuning	Updating documents/index is easier than retraining
Build model for unique domain with large proprietary dataset	Training or deep fine-tuning	Only when simpler options cannot meet requirements
Enforce deterministic business logic	Tool/function call or rules engine	LLM text generation is probabilistic
Execute workflow across systems	Agent with tools and guardrails	Requires permissions, logging, and human approval for risky actions

Prompt Engineering Patterns

Pattern	Use	Example instruction
Role	Set context	“Act as a network operations assistant.”
Task	Specify required output	“Summarize the incident in five bullets.”
Constraints	Limit scope	“Use only the provided context.”
Format	Make output parseable	“Return JSON with fields: severity, cause, next_action.”
Few-shot examples	Show desired behavior	Provide examples of input and output
Chain-of-thought alternative	Ask for concise rationale or steps without exposing hidden reasoning	“Provide a brief justification.”
Grounding	Tie answer to sources	“Cite the source chunk ID for each claim.”
Refusal rule	Prevent unsafe output	“If context is insufficient, say you do not know.”

Key Generative AI Parameters

Parameter	Effect	Exam note
Temperature	Higher values increase randomness	Use lower values for consistency
Top-p	Samples from most probable token mass	Another creativity/control setting
Max tokens	Caps generated output length	Prevents runaway responses but may truncate
System message	High-priority behavioral instruction	Useful for policy and role constraints
Stop sequence	Ends generation at defined text	Helpful for structured outputs
Context length	Amount of input/output the model can process	More context can increase cost and latency

Agentic AI Reference

Component	Function	Control requirement
Planner	Breaks goal into steps	Limit scope and validate plans
Tools/functions	APIs, databases, scripts, ticketing, network automation	Least privilege and allowlists
Memory	Stores prior context or state	Retention and privacy controls
Orchestrator	Coordinates model, tools, and state	Logging and error handling
Human-in-the-loop	Human approval for sensitive actions	Required for high-impact changes
Guardrails	Policy, safety, schema, and action constraints	Test with adversarial inputs
Audit trail	Records prompts, tool calls, outputs, and approvals	Needed for troubleshooting and accountability

Exam trap: An agent that can call tools is not just a chatbot. It becomes an automation system and must be governed like one.

Cisco-Oriented Architecture Decision Points

In Cisco-focused scenarios, map AI requirements to network, security, observability, collaboration, and data center design. The exam may describe AI in a branch, campus, data center, cloud, security operations, contact center, or network operations context.

Domain	AI use cases	Design priorities	Watch for
Campus/branch	Local inference, smart cameras, user assistance, operational analytics	Latency, segmentation, device identity, bandwidth, physical constraints	Do not send sensitive data to cloud by default
Data center	Model hosting, inference clusters, training, vector databases, data pipelines	High-throughput fabric, east-west traffic, storage performance, telemetry	AI traffic can be bursty and bandwidth-intensive
Cloud/hybrid	Managed AI services, burst compute, SaaS integrations	Secure connectivity, identity, data residency, cost visibility	Cloud reduces operations burden but not governance responsibility
Edge	Low-latency decisions, disconnected operation, local privacy	Small models, accelerators, lifecycle management	Edge devices may have limited compute and update windows
Security operations	Alert enrichment, anomaly detection, triage, summarization	Explainability, audit logs, false positive control	AI can assist analysts but should not blindly auto-remediate high-risk events
Network operations	AIOps, telemetry correlation, root-cause assistance, predictive maintenance	Data quality, baselines, topology context, observability	Correlation is not causation
Collaboration/contact center	Transcription, summarization, virtual agents, sentiment, routing	PII handling, user consent where applicable, quality monitoring	Generated summaries must be validated for accuracy
Application operations	Code assistance, log analysis, incident summaries	Secure SDLC, secrets handling, prompt/data controls	Do not expose credentials or proprietary code without controls

Infrastructure for AI Workloads

Component	Training priority	Inference priority	Exam distinction
CPU	Data preprocessing, orchestration, lightweight models	Low-volume or simple inference	General purpose but slower for many deep learning tasks
GPU/accelerator	Parallel matrix operations, model training	High-throughput inference	Critical for deep learning performance
Memory	Large batches, model parameters, feature sets	Model size and context handling	Memory pressure causes failures or latency
Storage	Large datasets, checkpoints, artifacts	Model files, embeddings, logs	Throughput and data locality matter
Network	Distributed training, data movement, storage access	API calls, service-to-service traffic	AI clusters can generate heavy east-west traffic
Observability	Job metrics, utilization, failures	Latency, errors, quality, drift	Monitor both infrastructure and model behavior

Notes and examples

Placement Decisions

Requirement	Prefer	Reason
Lowest latency near data source	Edge/local inference	Reduces round-trip time
Sensitive data should remain local	Edge or private data center	Limits data exposure
Large-scale model training	Data center or cloud with accelerators	Needs compute, storage, and high-speed networking
Variable demand	Cloud or elastic hybrid design	Scales with workload
Disconnected operation	Edge	Continues without reliable WAN
Centralized governance and shared services	Data center or cloud	Easier policy and lifecycle management
Cost-sensitive predictable workload	Compare on-premises/private vs cloud	Utilization and operations model matter

Networking Considerations for AI

Concern	Why it matters	Design response
East-west traffic	Distributed training and microservices exchange high volumes internally	Use high-throughput, low-latency fabric design
North-south traffic	Users, APIs, and cloud services access AI applications	Secure ingress/egress and policy enforcement
Latency	Affects interactive inference and real-time automation	Place inference near users/data when needed
Bandwidth	Large datasets, embeddings, and model artifacts are heavy	Plan data movement and storage locality
Jitter/packet loss	Can affect streaming analytics and interactive services	Use QoS and resilient paths where appropriate
Segmentation	AI systems may touch sensitive data and tools	Isolate workloads and enforce least privilege
Telemetry	AI operations need logs, metrics, traces, and network context	Collect end-to-end visibility
API connectivity	Agents and AI apps call tools/services	Secure APIs with authentication, authorization, and rate controls

Security and Governance

AI Risk-to-Control Matrix

Risk	Description	Controls
Prompt injection	User or document instructions manipulate model behavior	Input filtering, context isolation, tool allowlists, instruction hierarchy
Data leakage	Sensitive data appears in prompts, logs, outputs, or training data	Data classification, masking, encryption, access control, retention limits
Model hallucination	Unsupported or false output	RAG, citations, refusal rules, evaluation, human review
Data poisoning	Training or retrieval data is maliciously altered	Source validation, integrity checks, approval workflows
Model theft	Unauthorized access to model weights or endpoints	Strong IAM, network segmentation, monitoring
Adversarial input	Crafted input causes incorrect output	Robust testing, anomaly detection, guardrails
Bias/fairness issue	Outputs disadvantage groups or users	Representative data, bias testing, review process
Insecure tool use	Agent calls risky APIs or changes systems incorrectly	Least privilege, scoped tokens, human approval, audit logs
Over-permissioned service account	AI service has more access than needed	Role-based access control and periodic review
Unlogged decisions	No audit trail for AI-assisted actions	Prompt/output/tool-call logging with privacy controls
Shadow AI	Unapproved tools used with enterprise data	Policy, approved platforms, monitoring, user education

Notes and examples

Security Principles to Apply

Principle	AI-specific application
Least privilege	Models, agents, vector stores, and pipelines get only required access
Zero Trust	Verify users, devices, services, and context before access
Defense in depth	Combine IAM, segmentation, encryption, logging, and guardrails
Secure by design	Build controls into the AI workflow, not after deployment
Human oversight	Require approval for high-impact, irreversible, or risky actions
Data minimization	Use only data required for the task
Separation of duties	Separate model development, approval, deployment, and monitoring roles
Continuous monitoring	Watch quality, safety, security, and infrastructure signals

MLOps and LLMOps

Practice	ML focus	LLM/GenAI focus
Versioning	Data, features, code, model artifacts	Prompts, model versions, retrieval corpora, vector indexes
Registry	Approved models and metadata	Approved models, prompt templates, guardrails
CI/CD	Test and deploy application/model pipeline	Test prompts, eval sets, safety checks, integrations
Monitoring	Accuracy, drift, latency, resource use	Groundedness, hallucination rate, toxicity, retrieval quality, latency
Rollback	Revert model/application version	Revert prompt, model, index, or guardrail version
Experiment tracking	Hyperparameters and metrics	Prompts, model settings, retrieval parameters, eval results
Approval workflow	Promote model to production	Approve model, data sources, tools, and policies
Incident response	Model failure or degraded quality	Unsafe output, data exposure, prompt injection, tool misuse

Notes and examples

Production Readiness Checklist

Defined business owner and technical owner.
Documented data sources, sensitivity, and allowed use.
Baseline model or non-AI alternative compared.
Evaluation metrics aligned to business risk.
Security controls for identity, network, data, and APIs.
Logging for prompts, responses, tool calls, and errors where appropriate.
Monitoring for latency, cost, quality, drift, and abuse.
Rollback plan for model, prompt, retrieval index, or application release.
Human review path for high-impact decisions.
User communication that AI output may need verification.

What must be versioned

Artifact	Why version it
Dataset	Reproduce training and investigations.
Labels	Track changes in ground truth.
Feature definitions	Prevent silent changes in model inputs.
Code	Rebuild the pipeline.
Model artifact	Know exactly what is deployed.
Hyperparameters	Reproduce training behavior.
Prompt templates	Control LLM behavior changes.
Retrieval index	Trace RAG answer sources.
Evaluation set	Compare versions consistently.
Configuration	Reproduce deployment and runtime behavior.

Deployment patterns

Pattern	Use case	Watch for
Batch inference	Periodic scoring, reports, offline recommendations.	Freshness and scheduling.
Real-time inference	User-facing predictions or chat.	Latency, availability, scaling.
Streaming inference	Continuous data such as telemetry or sensors.	Throughput, ordering, backpressure.
Canary deployment	Release to small percentage first.	Monitoring and rollback criteria.
Blue-green deployment	Switch between old and new environments.	Compatibility and cutover control.
A/B testing	Compare model variants.	Statistical validity and user impact.
Human-in-the-loop	High-risk or ambiguous decisions.	Workflow design and escalation.

Monitoring after deployment

Monitor	Why it matters
Input data quality	Detect missing, invalid, or shifted inputs.
Prediction distribution	Identify unexpected output patterns.
Latency and errors	Maintain service reliability.
Resource utilization	Control cost and performance.
Business outcomes	Confirm model value.
Ground truth performance	Measure accuracy when labels arrive.
Drift	Decide whether retraining is needed.
Safety events	Detect harmful outputs or policy violations.
User feedback	Capture quality issues not visible in metrics.

Drift distinctions

Type	Meaning	Example
Data drift	Input distribution changes.	User behavior shifts after a new product launch.
Concept drift	Relationship between inputs and target changes.	Fraud patterns change after attackers adapt.
Prediction drift	Output distribution changes.	Model suddenly predicts one class more often.
Performance drift	Actual measured accuracy or usefulness declines.	Validation-like performance no longer matches production.

Exam decision rule: If inputs changed, think data drift. If the meaning of inputs changed relative to outcomes, think concept drift. If outputs changed, think prediction drift. If measured quality dropped, think performance drift.

Compact Code and Workflow Snippets

Basic ML Evaluation Pattern

## Conceptual pattern: split, train, validate, test
X_train, X_test, y_train, y_test = split_data(X, y, stratify=y)

model.fit(X_train, y_train)
predictions = model.predict(X_test)

print(classification_report(y_test, predictions))

Exam points:

Use stratified splitting for imbalanced classification when appropriate.
Do not train on the test set.
For time-series data, split by time rather than random order.

RAG Application Pattern

question = "What is the recommended remediation for this alert?"

query_vector = embed(question)
chunks = vector_index.search(query_vector, top_k=5, filters={"source": "approved"})
prompt = build_prompt(question=question, context=chunks, rule="Answer only from context.")

answer = llm.generate(prompt, temperature=0)
log_interaction(question, chunks, answer)

Retrieval quality is as important as generation quality.
Use metadata filters for source, tenant, sensitivity, and freshness.
Low temperature improves consistency but does not guarantee correctness.

Agent Tool-Use Pattern

if user_is_authorized and action in approved_actions:
    plan = agent.create_plan(request)
    if plan.requires_change:
        require_human_approval(plan)
    result = agent.call_tool(plan.tool, scoped_token)
    audit_log(plan, result)
else:
    deny_request()

Tool access requires authorization separate from chat access.
High-impact actions should use approval gates.
Audit the plan, tool call, result, and user identity.

Troubleshooting AI Systems

Symptom	Likely area	What to check
High inference latency	Infrastructure/model/API	Model size, accelerator use, network path, batching, token count
Poor answer quality	Data/model/prompt	Prompt clarity, source quality, evaluation set, retrieval quality
RAG answers cite irrelevant content	Retrieval	Chunking, embeddings, metadata filters, reranking
RAG says “not found” for known content	Indexing	Document ingestion, freshness, permissions, top-k, chunk size
Sudden model performance drop	Drift or data pipeline	Input distribution, schema changes, new user behavior
Many false positives	Threshold/metric mismatch	Precision, threshold, class balance, business cost
Many false negatives	Recall issue	Threshold, training labels, minority-class representation
Biased or unfair outputs	Data/model governance	Dataset representation, label bias, evaluation by subgroup
GPU underutilization	Pipeline bottleneck	Data loading, batch size, CPU preprocessing, network/storage
Excessive cost	Usage/model design	Token count, model size, caching, batching, placement
Unsafe or policy-violating output	Guardrails/security	System prompt, content filters, jailbreak tests, logging
Agent performs wrong action	Tool governance	Permissions, tool schema, approval flow, validation

Common Exam Traps

Trap	Correct exam thinking
“Use AI for everything”	Choose rules or automation when deterministic logic is enough
“Accuracy is the best metric”	Match metric to risk, especially with imbalanced classes
“More data always improves the model”	Quality, relevance, labeling, and leakage matter
“Fine-tuning adds current knowledge”	RAG is usually better for changing enterprise facts
“RAG eliminates hallucinations”	RAG reduces risk but still needs grounding and evaluation
“Cloud means no security responsibility”	Identity, data protection, monitoring, and governance still apply
“Training and inference are the same”	Training learns parameters; inference uses the model
“LLMs are deterministic search engines”	LLMs generate probabilistic outputs
“Unsupervised learning uses labeled examples”	Unsupervised learning finds patterns without labels
“Correlation proves root cause”	AIOps findings need validation and context
“Agents are just prompts”	Agents can take actions and require tool security
“Encryption fixes AI safety”	Encryption protects data in transit/at rest; it does not validate output
“Test set can be reused for tuning”	Use validation for tuning and test for final unbiased evaluation
“Edge is always better for privacy”	Edge helps data locality but still needs lifecycle, access, and logging controls

Cisco Scenario Checklist

Use this checklist when a question describes an AI solution in a Cisco environment:

Identify the outcome
- Prediction, classification, anomaly detection, summarization, Q&A, automation, or generation.
Classify the data
- Structured, telemetry, logs, documents, audio, video, sensitive, regulated, or proprietary.
Choose the AI pattern
- Rules, supervised ML, unsupervised ML, deep learning, LLM, RAG, or agentic workflow.
Place the workload
- Edge for low latency/locality.
- Data center for controlled high-performance hosting.
- Cloud for managed services and elasticity.
- Hybrid when data, latency, or governance requirements are mixed.
Design the network
- Account for bandwidth, latency, segmentation, API paths, telemetry, and resilience.
Apply security
- Identity, least privilege, encryption, segmentation, prompt/data protection, and audit logging.
Plan operations
- Model/prompt versioning, monitoring, incident response, rollback, and lifecycle management.
Validate the answer
- Prefer the option that aligns AI method, data, infrastructure, security, and business risk.

Final Review Priorities

Before sitting for Cisco AI Technical Practitioner (810-110 AITECH), be able to quickly explain:

Difference between AI, ML, deep learning, generative AI, RAG, and agents.
When to use supervised, unsupervised, reinforcement, and deep learning methods.
Why data quality, leakage, labeling, and bias affect model outcomes.
How to choose metrics based on false positive and false negative costs.
How RAG works and when it is better than fine-tuning.
How prompt engineering, model parameters, and guardrails affect LLM behavior.
How AI placement affects latency, privacy, cost, network design, and operations.
How Cisco-style architectures connect AI with networking, security, observability, and automation.
How to troubleshoot AI systems across model, data, prompt, retrieval, infrastructure, and security layers.

High-yield review map

Area	What to know quickly	Common exam trap
AI vs ML vs deep learning vs GenAI	AI is the broad field; ML learns patterns from data; deep learning uses neural networks; GenAI creates new content from learned patterns.	Treating every AI solution as generative AI or assuming all AI requires deep learning.
Data lifecycle	Collection, labeling, cleaning, splitting, feature engineering, validation, lineage, privacy, and monitoring.	Ignoring data quality and jumping directly to model choice.
Model types	Supervised, unsupervised, reinforcement learning, classical ML, neural networks, LLMs, embeddings.	Choosing a model type without checking labels, target output, latency, explainability, and data volume.
Evaluation	Classification, regression, clustering, ranking, GenAI quality, business impact, latency, and cost.	Optimizing one metric while missing the operational goal.
GenAI and LLMs	Prompts, tokens, context windows, embeddings, RAG, fine-tuning, hallucination controls, guardrails.	Assuming RAG, fine-tuning, and prompt engineering solve the same problem.
Infrastructure	Compute, GPU/accelerator use, memory, storage, network throughput, latency, scalability, resilience.	Designing for training needs when the workload is actually inference, or the reverse.
Networking for AI	East-west traffic, bandwidth, congestion, telemetry, segmentation, secure connectivity, edge/cloud placement.	Forgetting that AI performance can be limited by network and storage, not only compute.
Security and governance	Data protection, identity, access control, model risk, prompt injection, supply chain, auditability.	Applying only traditional app security and missing AI-specific threats.
MLOps/LLMOps	Versioning, deployment, model registry, monitoring, drift detection, rollback, retraining, evaluation gates.	Thinking deployment is the end of the AI lifecycle.

Core AI concepts candidates should distinguish

AI, ML, deep learning, and generative AI

Concept	Plain-language meaning	Typical output
Artificial intelligence	Systems that perform tasks associated with human-like reasoning, perception, language, or decision-making.	Prediction, recommendation, classification, generated content, automation.
Machine learning	Systems learn patterns from data rather than being explicitly programmed for every rule.	Class label, numeric prediction, ranking, cluster, anomaly score.
Deep learning	ML using multi-layer neural networks, often effective for images, speech, language, and large-scale pattern recognition.	Image labels, transcriptions, embeddings, generated text, detections.
Generative AI	Models that create new text, images, audio, code, or other content.	Drafts, summaries, chat responses, synthetic media, generated code.
Foundation model	Large pre-trained model adaptable to many downstream tasks.	General language, vision, or multimodal outputs.
LLM	Large language model specialized for text and language-like token prediction.	Text, code, structured responses, reasoning-like outputs.

Notes and examples

Exam decision rule: If the question asks for a prediction from labeled historical examples, think supervised ML. If it asks for grouping without labels, think unsupervised learning. If it asks for content generation, summarization, conversation, or transformation, think GenAI/LLM. If it asks for interaction with an environment and rewards, think reinforcement learning.

Evaluation metrics that show up in practice

Classification metrics

Metric	Plain formula	Use when…	Trap
Accuracy	Correct predictions / total predictions	Classes are balanced and all errors have similar cost.	Misleading on imbalanced datasets.
Precision	TP / (TP + FP)	False positives are costly.	High precision can miss many true cases.
Recall / sensitivity	TP / (TP + FN)	False negatives are costly.	High recall may increase false positives.
F1 score	Harmonic mean of precision and recall	Need balance between precision and recall.	Hides the business meaning of each error type.
Specificity	TN / (TN + FP)	Need to measure true negative rate.	Not enough by itself for rare positive classes.
ROC-AUC	Ranking quality across thresholds	General binary classifier comparison.	Can look strong even when rare-event performance is weak.
PR-AUC	Precision-recall tradeoff	Imbalanced positive class.	Harder to interpret without baseline prevalence.

Notes and examples

Regression and forecasting metrics

Metric	What it emphasizes	Trap
MAE	Average absolute error; easy to interpret.	Does not penalize large errors as heavily as RMSE.
RMSE	Larger errors more heavily penalized.	Sensitive to outliers.
R-squared	Proportion of variance explained.	Can look acceptable while business error is too high.
MAPE	Percentage error.	Breaks down or becomes unstable near zero values.

GenAI and LLM evaluation

Evaluation dimension	What to check
Correctness	Is the answer factually right for the context?
Groundedness	Is the answer supported by provided or retrieved sources?
Relevance	Does it answer the user’s actual request?
Safety	Does it avoid harmful, sensitive, or prohibited output?
Robustness	Does it resist prompt injection, ambiguity, and adversarial inputs?
Consistency	Does it provide stable behavior across similar prompts?
Latency	Does it respond quickly enough for the use case?
Cost	Are token, compute, and infrastructure costs sustainable?
Explainability	Can the system provide traceable rationale or citations where needed?

Exam trap: BLEU, ROUGE, or automated similarity metrics may help for some language tasks, but they do not fully prove that an LLM answer is correct, safe, or grounded.

Generative AI and LLM review

Prompt engineering essentials

A strong prompt often includes:

Role or task — what the model should do.
Context — facts, documents, constraints, or user background.
Instructions — steps, rules, and boundaries.
Examples — desired input/output patterns.
Output format — table, JSON, bullets, summary, classification label.
Safety constraints — what not to reveal or perform.
Success criteria — what a good answer should satisfy.

Prompting technique	Use case	Trap
Zero-shot	Direct task with no examples.	May be inconsistent for nuanced outputs.
Few-shot	Provide examples of desired behavior.	Bad examples can anchor bad behavior.
Chain-of-thought style guidance	Encourage structured reasoning or stepwise analysis.	For production, prefer concise rationale or verifiable steps rather than exposing unnecessary internal reasoning.
Structured output	Need parseable JSON, tables, or labels.	Must validate output; models can still produce malformed structures.
System instructions	Set higher-priority behavior and boundaries.	Not a complete security control by itself.

Notes and examples

Tokens, context, temperature, and output control

Concept	Meaning	Practical effect
Token	Unit of text processed by the model.	Drives context size, cost, and latency.
Context window	Maximum tokens the model can consider at once.	Long documents may need chunking or retrieval.
Temperature	Controls randomness.	Lower for deterministic tasks; higher for creative variation.
Top-p	Controls probability mass considered for generation.	Another way to tune output diversity.
Max tokens	Output length limit.	Too low truncates answers; too high increases cost.
Stop sequence	Pattern that ends generation.	Useful for structured outputs or agent boundaries.

RAG, fine-tuning, and prompt-only approaches

Need	Best starting approach	Why
Answer using current private documents	Retrieval-augmented generation	Keeps knowledge external and updateable.
Improve response style or format	Prompting or fine-tuning	Depends on consistency need and volume.
Teach new facts that change often	RAG	Updating an index is easier than retraining.
Specialize behavior across many examples	Fine-tuning	Useful when repeated prompt examples are not enough.
Reduce hallucinations from missing context	RAG plus grounding checks	The model needs access to trusted information.
Build from proprietary domain data at scale	Fine-tuning or custom training	Requires governance, data quality, and infrastructure.

RAG pipeline essentials

Step	Purpose	Common issue
Document ingestion	Bring source content into the system.	Untrusted, stale, or duplicate content.
Chunking	Split documents into retrievable pieces.	Chunks too small lose context; chunks too large reduce precision.
Embedding	Convert text into vectors for similarity search.	Poor embedding model for domain language.
Vector search	Retrieve semantically related chunks.	Retrieves similar but not authoritative content.
Reranking	Improve result ordering.	Adds latency but can improve relevance.
Prompt assembly	Combine user question and retrieved context.	Context overflow or irrelevant context.
Generation	Produce final answer.	Hallucination, overconfidence, missing citations.
Evaluation	Measure correctness and groundedness.	Relying only on user satisfaction.

Decision rule: Use RAG when the model needs trusted, updateable, external knowledge. Use fine-tuning when the model needs consistent behavior, tone, format, or task adaptation that prompts cannot reliably achieve.

AI infrastructure and networking review

For a Cisco exam, connect AI concepts to technical infrastructure: where data moves, how workloads scale, how systems are secured, and how networks support high-throughput, low-latency operations.

Training vs inference

Dimension	Training	Inference
Goal	Learn model parameters from data.	Use a trained model to produce outputs.
Compute pattern	Heavy, often distributed, accelerator-intensive.	Latency-sensitive, may need autoscaling.
Data flow	Large datasets, checkpoints, repeated reads/writes.	Requests and responses, sometimes retrieval calls.
Network concern	High east-west traffic between nodes; synchronization.	User latency, API throughput, availability.
Storage concern	Dataset access, checkpointing, versioning.	Model loading, cache, retrieval index access.
Optimization	Throughput, utilization, parallelism.	Latency, concurrency, cost per request.
Failure impact	Lost training time, checkpoint recovery.	User-facing outage or degraded service.

Notes and examples

Infrastructure bottlenecks

Symptom	Likely area to investigate
GPUs underutilized during training	Data pipeline, storage throughput, network bottleneck, small batch size.
Training fails intermittently	Node failure, network instability, memory exhaustion, dependency mismatch.
Inference latency is high	Model size, cold starts, retrieval delay, network path, overloaded service.
RAG answers are slow	Vector database latency, reranking, large context, too many retrieval calls.
Model loads slowly	Storage performance, image size, model artifact location.
Cost spikes	Overprovisioned accelerators, excessive tokens, inefficient batching, poor autoscaling.
Production accuracy drops	Data drift, concept drift, upstream data changes, label delay.

Networking concepts for AI workloads

Concept	Why it matters
Bandwidth	Large datasets, distributed training, and model artifacts can move significant traffic.
Latency	Affects user-facing inference, API chains, RAG retrieval, and distributed synchronization.
East-west traffic	AI clusters often communicate heavily between compute nodes.
Congestion management	Prevents throughput collapse under high load.
Segmentation	Isolates sensitive data, model services, management planes, and tenants.
Telemetry	Helps detect bottlenecks, errors, drops, latency, and capacity issues.
Resilience	AI services need redundancy, failover, and graceful degradation.
Edge placement	Reduces latency and bandwidth use when data is generated near users or devices.
Secure connectivity	Protects data in transit and controls access to models and APIs.

Cloud, on-premises, hybrid, and edge placement

Placement	Strengths	Tradeoffs
Public cloud	Elastic capacity, managed AI services, rapid experimentation.	Cost control, data residency, egress, shared responsibility complexity.
On-premises data center	Control, proximity to sensitive data, predictable governance.	Capacity planning, hardware cost, operational responsibility.
Hybrid	Balance control with cloud flexibility.	Integration, identity, network, and policy consistency.
Edge	Low latency, local autonomy, reduced backhaul.	Limited compute, lifecycle management, physical security.

Exam decision rule: If the requirement emphasizes real-time response near devices, consider edge inference. If it emphasizes massive training capacity and elasticity, cloud may fit. If it emphasizes data control, compliance, or existing private infrastructure, on-premises or hybrid may be favored.

Security, privacy, and responsible AI

Traditional security still applies

AI systems still need normal enterprise controls:

Strong identity and access management.
Least privilege.
Encryption in transit and at rest.
Network segmentation.
Secure APIs.
Logging and monitoring.
Vulnerability management.
Backup and recovery.
Supply chain control.
Incident response.

Notes and examples

AI-specific risks

Risk	What it means	Mitigation direction
Prompt injection	User or retrieved content attempts to override instructions.	Input isolation, instruction hierarchy, content filtering, tool-use controls.
Data leakage	Sensitive information appears in prompts, logs, training data, or outputs.	Data classification, redaction, access control, retention limits.
Model inversion	Attacker infers training data from model behavior.	Privacy controls, output limits, careful training data governance.
Model theft	Unauthorized copying or extraction of model behavior.	API rate limits, monitoring, access controls, watermarking where appropriate.
Poisoning	Malicious data affects training or retrieval.	Data validation, trusted sources, lineage, anomaly detection.
Hallucination	Model produces plausible but false content.	RAG, citations, confidence handling, human review for high-risk tasks.
Bias and unfairness	Model performance differs across groups.	Representative data, fairness testing, monitoring, governance review.
Unsafe tool use	Agent or model takes harmful action through connected tools.	Permission boundaries, approvals, sandboxing, audit logs.
Shadow AI	Unapproved AI use with sensitive data.	Policy, approved tools, monitoring, user education.

Responsible AI review points

Principle	Practical meaning
Transparency	Users and stakeholders understand AI involvement and limitations.
Accountability	Owners are defined for model behavior, data, and incidents.
Fairness	Systems are tested for harmful bias and unequal impact.
Privacy	Personal and sensitive data is protected throughout the lifecycle.
Safety	Outputs and actions are controlled for harmful outcomes.
Reliability	System behavior is monitored and validated over time.
Explainability	Decisions can be understood at the level required by the use case.

Common trap: Responsible AI is not only a documentation exercise. It affects data selection, model evaluation, deployment controls, user experience, monitoring, and incident response.

Common scenario decisions

Which metric should be prioritized?

Scenario	Better metric focus
Detecting a dangerous condition where missing it is costly	Recall / sensitivity
Alerting analysts where too many false alarms waste time	Precision
Balanced classification with similar error costs	Accuracy may be acceptable
Rare-event detection	Precision, recall, F1, PR-AUC
Numeric forecast with large errors especially harmful	RMSE
Numeric forecast needing easy business interpretation	MAE
LLM answers for internal knowledge base	Groundedness, correctness, citation quality, latency
AI assistant with connected tools	Safety, authorization, auditability, task success

Notes and examples

Which control reduces which GenAI risk?

Requirement	Strong control
Keep answers tied to company documents	RAG with trusted sources and citations.
Prevent sensitive data exposure	Data loss prevention, redaction, access controls, logging policy.
Stop model from taking unauthorized actions	Tool permission boundaries and approval workflows.
Reduce prompt injection impact	Treat retrieved/user content as untrusted, constrain tools, validate outputs.
Improve consistent response format	Structured prompts, schema validation, possibly fine-tuning.
Improve factual currency	Retrieval from updated sources, not static model memory.
Investigate bad answers	Trace prompts, model version, retrieved chunks, user context, logs.

Which infrastructure issue is most likely?

Clue in question	Likely answer direction
“High GPU cost but low utilization”	Data/input pipeline or scheduling inefficiency.
“Model works in lab but fails with live traffic”	Production data mismatch, scaling, latency, drift, or integration issue.
“Private data cannot leave environment”	On-premises, private cloud, hybrid controls, or local inference.
“Need milliseconds of response near devices”	Edge inference or local processing.
“Need to query changing internal documents”	RAG and document indexing.
“Responses cite wrong or stale policy”	Retrieval source quality, index freshness, chunking, or ranking.
“Users can make the assistant ignore instructions”	Prompt injection and insufficient guardrails.
“Accuracy high but minority class missed”	Class imbalance; use recall, precision, F1, PR-AUC.

Candidate mistakes to avoid

Confusing training and inference. Training creates or updates the model; inference uses it.
Using accuracy by default. Accuracy can be the wrong metric for imbalanced or high-risk cases.
Ignoring data leakage. Leakage creates impressive but false performance.
Assuming bigger models are always better. Larger models can increase cost, latency, risk, and complexity.
Treating prompts as security boundaries. Prompts guide behavior but do not replace access control and validation.
Choosing fine-tuning when RAG is the better answer. New or private facts usually belong in retrieval, not model weights.
Forgetting monitoring. AI behavior can degrade after deployment.
Overlooking network and storage bottlenecks. AI workloads are not only compute problems.
Ignoring governance. Data lineage, auditability, privacy, and responsible AI are part of production readiness.
Missing the business goal. The best technical metric may not match the operational requirement.

Fast final-review checklist

Before you move into practice questions, make sure you can answer these without notes:

Can you explain AI, ML, deep learning, GenAI, LLMs, and foundation models distinctly?
Can you choose supervised, unsupervised, reinforcement, RAG, or fine-tuning based on a scenario?
Can you identify data leakage, drift, imbalance, and poor data splitting?
Can you choose precision, recall, F1, MAE, RMSE, or groundedness for the right situation?
Can you explain why training and inference have different infrastructure needs?
Can you identify when latency, bandwidth, storage, or compute is the bottleneck?
Can you describe the role of embeddings and vector search in RAG?
Can you recognize hallucination, prompt injection, poisoning, and data leakage risks?
Can you outline a basic MLOps/LLMOps lifecycle from data to monitoring?
Can you connect AI workloads to secure, resilient network design?

Put the review into practice

Study Plan

810-110 AITECH — Cisco AI Technical Practitioner Cheat Sheet

AI, ML, and Generative AI Terms

AI Workload Lifecycle

Choosing the Right AI Approach

Model and Algorithm Selection

Model selection quick rules

Classical ML vs deep learning vs LLMs

Data Reference

Data Types

Data Preparation Decisions

Data lifecycle review

Data concepts to review

Data splitting trap

Evaluation Metrics

Confusion Matrix Terms

Generative AI and RAG

RAG Pipeline

Prompting, RAG, Fine-Tuning, or Training?

Prompt Engineering Patterns

Key Generative AI Parameters

Agentic AI Reference

Cisco-Oriented Architecture Decision Points

Infrastructure for AI Workloads

Placement Decisions

Networking Considerations for AI

Security and Governance

AI Risk-to-Control Matrix

Security Principles to Apply

MLOps and LLMOps

Production Readiness Checklist

What must be versioned

Deployment patterns

Monitoring after deployment

Drift distinctions

Compact Code and Workflow Snippets

Basic ML Evaluation Pattern

RAG Application Pattern

Agent Tool-Use Pattern

Troubleshooting AI Systems

Common Exam Traps

Cisco Scenario Checklist

Final Review Priorities

High-yield review map

Core AI concepts candidates should distinguish

AI, ML, deep learning, and generative AI

Evaluation metrics that show up in practice

Classification metrics

Regression and forecasting metrics

GenAI and LLM evaluation

Generative AI and LLM review

Prompt engineering essentials

Tokens, context, temperature, and output control

RAG, fine-tuning, and prompt-only approaches

RAG pipeline essentials

AI infrastructure and networking review

Training vs inference

Infrastructure bottlenecks

Networking concepts for AI workloads

Cloud, on-premises, hybrid, and edge placement

Security, privacy, and responsible AI

Traditional security still applies

AI-specific risks

Responsible AI review points

Common scenario decisions

Which metric should be prioritized?

Which control reduces which GenAI risk?

Which infrastructure issue is most likely?

Candidate mistakes to avoid

Fast final-review checklist

Put the review into practice

Browse Certification Practice Tests